Expected utility for nonstochastic risk

The world of random phenomena exceeds the domain of the classical probability theory. In the general case the description of randomness requires a specific set of probability distributions (which is called statistical regularity) rather than a singe distribution. Such statistical regularity arises as a limit of relative frequencies. This approach to randomness allows to generalize the expected utility theory in order to cover the decision problems under nonstochastic random events. Applying the von Neumann-Morgenstern utility theorem, we derive the maxmin expected utility representation for statistical regularities. The derivation is based on the axiom of the preference for stochastic risk, i.e. the decision maker wishes to reduce the set of probability distributions to a single one.